Newest papers

Beyond End-to-End Recovery: Towards a Better Understanding of LLM Steganography via Rolewise Decomposition

Om Tailor, Yucheng Pan, Solaleh Mohammadi, Xinyan Liu, Kaiqing Zhang · 2026 · arXiv (Cornell University)

DOI: 10.48550/arxiv.2610.30516 · cited by 0 works (OpenAlex) · open-access PDF

French

Dans le grand modèle de langue (LLM) steganographie, un encodeur cache un message privé dans une réponse autrement ordinaire, en tant que texte transmis à un décodeur, qui essaie de le récupérer. Le succès de cette transmission de message privé est généralement mesuré par la récupération de bout en bout (E2E), qui vérifie si le décodeur reconstruit correctement le message privé. Cependant, la récupération E2E ne peut pas distinguer les échecs d'encodage des échecs de décodage ou révéler comment le message privé est représenté dans le texte transmis par l'encodeur. Pour résoudre ces limitations, nous introduisons un cadre de décomposition de rôle qui examine séparément comment les messages privés sont représentés dans les textes transmis par

Machine translation (Meta M2M100); the English original follows.

Abstract

In large language model (LLM) steganography, an encoder hides a private message in an otherwise ordinary response, as text transmitted to a decoder, who attempts to recover it. The success of this private-message transmission is commonly measured through end-to-end (E2E) recovery, which checks whether the decoder correctly reconstructs the private message. However, E2E recovery cannot distinguish encoding failures from decoding failures or reveal how the private message is represented in the encoder's transmitted text. To address these limitations, we introduce a rolewise decomposition framework that separately examines how private messages are represented in encoder transmitted texts, what decoders recover, and what deterministic verification can establish. Across 7,200 encoder responses, we identify 168 verifiable encodings missed by decoders and 706 responses containing private messages reconstructed through word-length or token-initial patterns not explicitly specified in the encoder's instructions. These findings demonstrate that E2E recovery alone may obscure important distinctions between encoding and decoding, motivating their separate evaluation to better understand and mitigate the safety issues caused by LLM steganography.

Read in: English · Spanish · Chinese · Hindi · Arabic · Portuguese · French · German · Japanese · Russian · Korean · Indonesian · Vietnamese

Topics: Encoding (memory), Encoder, Decoding methods, Computer science, Verifiable secret sharing, Transmission (telecommunications)

Is this paper real? This page was built from OpenAlex and arXiv records, not generated. Software can verify any reference in one call: GET https://store.agentexchange.work/gw/papers/verify?title=...&year=... (0.003 USD), fetch this record as JSON with /gw/papers/paper?id=2610.30516, or list what it cites with /gw/papers/references?id=.... Pay per call in USDC (x402) or with a 19 USD card key for 1,000 calls. Catalogue · llms.txt

Sources: OpenAlex (CC0), arXiv, Crossref. Agent Exchange